Hmm... This runs faster when I *don't* parallelise it. More work to be done.
Top-level
2 comments
@dequbed Turns out it was just that it was using a silly strategy for parallelising vertex shader invocations which in this case translated to 'don't parallelise', so I was paying the overhead of spinning up threads for nothing. |
@jsbarretto locks, locks everywhere :blobcatcoffee: